Introduction

In modern agriculture, chemical herbicides play an essential role in efficient weed control. Meanwhile, the evolution of herbicide resistance in weeds has become a global problem seriously implicating the sustainability of arable agriculture [1, 2]. The mechanisms responsible for herbicide resistance can be classed into two categories: target-site-based resistance (TSR) and non-target-site-based resistance (NTSR) [3]. TSR is relatively easily analyzed, as most caused by single nucleotide polymorphisms in the “herbicide” target site encoding genes or overproduction of target site proteins [4,5,6]. Although the TSR accounts for a large majority of the reported herbicide resistance cases, chemical control can be readily achieved by alternating the use of herbicides with different modes of action (MOAs) [7]. In comparison, NTSR is a much less studied but more threatening mechanism. Most weeds harboring NTSR exhibit altered absorption, enhanced metabolism, or impaired translocation [8,9,10], counteracting herbicide-imposed toxicity irrespective of their MOAs. More and more studies have suggested the involvement of multiple genes in NTSR [11, 12], hampering the identification of genes involved in resistance. Therefore, only a few genes have been shown to be related to NTSR [13,14,15,16,17,18].

Among the NTSR mechanisms, enhanced herbicide metabolism, also named metabolic resistance, is a powerful route to come into being resistance to graminicides, as the differential rates of detoxification between grasses and cereals represent a key biochemical feature that can be exploited in selective chemical weed control [19]. For grass species, metabolic resistance is often associated with elevated levels of herbicide-detoxifying enzymes, including cytochrome P450 mixed-function oxidases (P450s), UDP-glucose-dependent glycosyltransferases (UGTs), glutathione S-transferases (GSTs), and membrane-associated ATP-binding cassette (ABC) drug transporter proteins [12, 20,21,22]. Of these, P450s are one of the largest superfamilies of enzyme proteins, found in the genomes of virtually all organisms [23]. Plant P450s are a group of heme-thiolate monooxygenases able to catalyze a wide variety of monooxygenation/hydroxylation reactions [24], participating in various biochemical pathways to produce primary and secondary metabolites [25]. A handful of P450s including CYP71, CYP72, CYP73, CYP76, CYP81, and CYP749 have been identified to deliver metabolic resistance to herbicides with different MOAs in field crop species [26]. A notable case reported recently is that CYP81As are demonstrated to be involved in the concomitant resistance of watergrass (Echinochloa phyllopogon) to multiple herbicides [14, 27, 28]. In most cases, while overexpression of specific P450s results in herbicide resistance, internal genetic mechanisms such as resistance origin and evolution remain unclear. To date, we are far from fully understanding metabolic herbicide resistance in weedy species due to its complexity of metabolic resistance and the diversity of plant P450s.

Shortawn foxtail (Alopecurus aequalis Sobol.) is a diploid (2n = 14) and partly cross-pollinated (c. 40%) species of the Poaceae family. This plant severely infests wheat (Triticum aestivum L.) and oilseed rape (Brassica napus L.) fields, causing significant reduction in grain yields. Mesosulfuron-methyl is a highly efficient acetolactate synthase (ALS)-inhibiting herbicide which has been frequently used for control of this weed in the last decade. In June 2013, herbicide-resistant A. aequalis was first found in Jiangsu Province, China [29]. Today, A. aequalis plants with TSR or both TSR and NTSR exhibit resistance to at least 10 herbicides from eight different chemical groups: aryloxyphenoxypropionate, cyclohexanedione, phenylpyraxoline, sulfonylurea, imidazolinone, triazolo-pyrimidine, pyrimidinyl-thiobenzoate, and sulfonyl-aminocarbonyl-triazolinone [30]. There seems to be, at least for now, only the photosystem II inhibitor isoproturon and the 4-hydroxyphenylpyruvate dioxygenase inhibitor cypyrafluone left to control this species in China. Investigating the resistance mechanisms in weeds is the base to delay or overcome herbicide resistance and to develop more effective pest management practices. Since the TSR has been well characterized in this species, elucidating the molecular mechanisms of NTSR is urgently required.

Previously, we have identified a field evolved metabolic resistant (R) A. aequalis population whose resistance to the ALS inhibitor mesosulfuron-methyl (MM) could be partly reversed by pre-treatment with the P450 inhibitor malathion [31]. In RNA sequencing, four P450s were identified with higher expression in the R vs. susceptible (S) lines [31]. Therefore, we hypothesize that rapid metabolism of MM results from increased activity of P450s, and that this increased activity results from higher expression of novel P450 isoforms with greater affinity towards MM. To date, involvement of P450s in metabolic herbicide resistance of A. aequalis has not been investigated in either biochemistry level or molecular level. Which P450 genes are responsible for and how the genes take part in the metabolic resistance still mysteries. Here we report the identification and functional characterization of key P450 genes endowing A. aequalis with resistance to ALS inhibitor(s). These results will provide a molecular basis for the understanding of the metabolic herbicide resistance in A. aequalis as well as other grass weeds in general.

Results

Metabolism of MM in the S and R lines of A. aequalis

According to Aizawa [32], the active body of MM would be primarily inactivated by O-demethylation after being absorbed by wheat (Fig. 1A). To determine if the metabolic pathway is the same as in wheat and if the MM metabolism is more active in the R than in the S line, the amounts of MM and its O-demethylated metabolite were detected by liquid chromatography coupled with tandem mass spectrometry (LC–MS/MS) and compared in the S and R lines of A. aequalis treated with MM for 24 h. In accordance with our previous results [33], the amount of MM in the R line was about three quarters of that in the S line, while that of O-demethylated metabolite was twofold higher (Fig. 1B). This reveals that MM was metabolized more rapidly in the R than in the S line via the same metabolic pathway as in wheat.

Fig. 1
figure 1

LC–MS/MS analyses of MM metabolism in S and R lines of A. aequalis. A Proposed metabolic pathway of MM in wheat. B Amounts of MM and its O-demethylated metabolite in the S (SD-01) and R (AH-18) lines of A. aequalis, respectively. Weed seedlings at the 3–4-leaf stage were treated with MM at the field recommended rate (9 g ai ha−1) for 24 h before LC–MS/MS analysis. *Significant difference at P < 0.05. NS no significant difference

Full size image

Full sequence cloning and analysis of the P450 genes

In previous study, RNA sequencing revealed that the expression of eight P450 genes was significantly higher in the R than in the S plants after MM treatment [31]. Of them, four P450 genes including CYP94C117 (c39857_g2), CYP709C56 (c45454_g1), CYP94E14 (c21190_g1), and CYP71R18 (c43350_g3) also showed consistently higher expression in parallel resistant samples than in susceptible samples (up to 32.64-fold) (Table S1). Considering that pre-treatment of the P450 inhibitor malathion greatly increased the susceptibility of the R plants to MM (by c. 41%) [31], we presumed the higher expression of these four P450 genes correlates with the enhanced MM metabolism in A. aequalis.

Full-length coding sequences of the four P450 genes were each amplified from the S and R plants of A. aequalis, and sequence alignments revealed no insertion or single nucleotide polymorphism was present for every gene between the two biotypes (Fig. S1 for CYP709C56; data not shown for the other three genes). Gene conserved domains were analyzed by the National Center for Biotechnology Information conserved domain tool (http://www.ncbi.nlm.nih.gov/structure/cdd/wrpsb.cgi), confirming the CYP94C117, CYP709C56, CYP94E14, and CYP71R18 all belong to the P450 superfamily. Sequence analysis of the above four A. aequalis genes and 219 P450 genes in Arabidopsis showed that CYP71R18 clustered in CYP71 clan, CYP709C56 clustered in CYP709 family of CYP72 clan, and CYP94C117 and CYP94E14 clustered in CYP94 family of CYP86 clan [34] (Fig. 2A). The nearest-neighbor analysis of characterized P450 protein sequences indicates that A. aequalis CYP709C56 has close evolutionary relationships with T. aestivum CYP709C1 and Aegilops tauschii CYP709B1, forming a sister clade with Brachypodium distachyon CYP709B1 (Fig. 2B).

Fig. 2
figure 2

Sequence analysis of A. aequalis P450 proteins. (1) Multiple sequence alignment of the A. aequalis CYP94C117, CYP709C56, CYP94E14, and CYP71R18 (black circle) and the 219 P450s from Arabidopsis species. (2) Phylogenetic analysis of the A. aequalis CYP709C56 and the CYP709s from other plant species including Brachypodium distachyon CYP709B1 (XP_003562608), Aegilops tauschii CYP709B1 (XP_020184145), Triticum aestivum CYP709C1 (AAT68297), Oryza sativa CYP709B1 (XP_015645350), Zea mays CYP709B2 (PWZ37282), Setaria italica CYP709B2 (XP_004958443), and Sorghum bicolor CYP709B2 (XP_002463247)

Full size image

Arabidopsis overexpressing CYP709C56 became less susceptible to specific ALS inhibitors

Full-length coding sequences of A. aequalis CYP94C117, CYP709C56, CYP94E14, and CYP71R18 were each introduced into Arabidopsis (ecotype Columbia-0) under the control of the cauliflower mosaic virus (CaMV) 35S promoter (Table S2). Arabidopsis seedlings overexpressing empty vector (pPZP211) were used as the negative control (WT). Ten independent T3 lines were obtained for each transformant, and they were first used for preliminary germination test on medium containing different classes of ALS herbicides.

Unexpectedly, the Arabidopsis lines overexpressing CYP94C117, CYP94E14, or CYP71R18 were as susceptible as WT, and their germinations were completely inhibited at 5 nM MM, 5 nM pyroxsulam (PX), or 15 μM flucarbazone–sodium (FS) (Fig. S2). Therefore, no further analysis of these genes was conducted. As expected, the seeds of Arabidopsis overexpressing CYP709C56 could still germinate at levels at which the WT stopped growing (Fig. 3A), indicating the gene endowed Arabidopsis with tolerance to MM and PX. However, both the WT and the Arabidopsis overexpressing CYP709C56 did not germinate at 15 μM FS (Fig. 3A), suggesting no tolerance of the transgenic lines occurred to FS. All following experiments were performed based on the CYP709C56 gene and the Arabidopsis lines overexpressing CYP709C56.

Fig. 3
figure 3

Susceptibility to ALS inhibitors in Arabidopsis transformed with CYP709C56. A Seedlings of the transgenic Arabidopsis lines CYP-1, CYP-2, and CYP-3 grown for 12 d on media containing no herbicide (CK), 5 nM mesosulfuron-methyl (MM), 5 nM pinoxaden (PX), or 15 μM flucarbazone–sodium (FS). B Transcript accumulation levels of T3 homozygous transgenic Arabidopsis lines CYP-1, CYP-2, and CYP-3. Different letters indicate a significant difference (P < 0.05). C MM susceptibility of independent transgenic lines was evaluated by whole-plant dose–response experiments. Bars represent SE (n = 3). D Relationship of transcript level and MM susceptibility in transgenic Arabidopsis overexpressing CYP709C56. Transcript level and MM susceptibility of each transgenic line were evaluated by RT-qPCR and GR50, respectively. Bars represent SE (n = 3). E Growth response to mesosulfuron-methyl of Arabidopsis transformed with the empty vector (WT) or CYP709C56 (CYP-1, CYP-2, and CYP-3). Growth status was checked at 9 days after treatment at the rate of 0 (CK) or 2.25 g ai ha−1 (T)

Full size image

Arabidopsis seedlings overexpressing CYP709C56 became resistant to mesosulfuron-methyl

To determine the relative expression levels of transgene in Arabidopsis, real-time quantitative PCR (RT-qPCR) assay was performed using the AtF-box protein as the internal control. The transgenic line exhibiting the lowest relative expression was defined as line CYP-10 with its transcript accumulation level was defined as 1.0. Three lines with higher transcript accumulation levels of transgene were selected for further physiological experiments.

Transcript accumulation levels of CYP709C56 were different among the three lines, which were separately marked as CYP-1 to CYP-3 based on their mean transcript levels from low to high (Fig. 3B). Whole-plant dose–response experiments were conducted to determine the susceptibility of all CYP709C56 transformants (Fig. 3C, E). MM susceptibility of each line differed, and the herbicide concentrations resulting in 50% growth reduction (GR50 values) of CYP-1 to CYP-3 ranged from 4.20 to 5.96 g active ingredient (ai) ha−1, which were 2.92- to 4.14-fold more resistant to MM than was control (Fig. 3C). The herbicide tolerance was positively related to the transcriptional level of the introduced CYP709C56 in the transgenic lines (Fig. 3D), suggesting CYP709C56 endows Arabidopsis with varying degrees of herbicide tolerance depending on its transcript abundance.

Arabidopsis overexpressing CYP709C56 metabolized mesosulfuron-methyl faster

To test if MM metabolism in the transgenic lines of CYP709C56 is more active than in the control plants, MM and its metabolites were examined in the T3 Arabidopsis lines overexpressing CYP709C56 or empty vector. Transgenic line CYP-3 with the highest transgenic expression level was used for the testing (Fig. 3B). LC–MS/MS analysis showed that herbicide content was not prominently different at 2 h (Fig. 4A), indicating the absorption of MM was similar in the WT and transgenic line [14]. As time goes on from 2 to 24 h, the amount of MM decreased rapidly in a similar way in both the WT and the transgenic line, and a significantly lower amount of MM was observed in the transgenic line (Fig. 4A). Similarly, the conversion of MM into non-herbicidal polar metabolite increased with time and peaked at 12 h in both the S and R lines, and the amount of O-demethylated MM was always significantly richer in R than that in S (P < 0.05) (Fig. 4B). However, the amount of MM metabolite decreased in both lines at 24 h (Fig. 4B), which may owe to their further degradation via enzyme systems in phase II and phase III of xenobiotic metabolism [35]. Clearly, overexpression of CYP709C56 increased the metabolism of MM to the O-demethylated metabolite, therefore, endowing plants with herbicide resistance.

Fig. 4
figure 4

Time course of MM metabolism in WT and transgenic lines of A. thaliana. Plants were foliar-treated with the field-recommended rate of MM at 9 g ai ha−1. The amounts of MM (A) and its O-demethylated metabolite (B) were monitored by LC–MS/MS at m/z 502.10 and 488.10, respectively. Bars represent SE (n = 3). *Significant difference at P < 0.05

Full size image

In vitro expressed CYP709C56 enzyme detoxifies MM via O-demethylation

To investigate and validate the function of CYP709C56, recombinant CYP709C56 protein was expressed as the N-terminal Flag-Tagged fusion proteins using a yeast (Saccharomyces cerevisiae) expression system that carried the Arabidopsis NADPH-cytochrome P450 reductase gene ATR1 [36]. The immunoblotting showed that CYP709C56 had accumulated in the transgenic yeast (Fig. 5A). The herbicide MM was added to the yeast culture media for the metabolism assay. After shaking for 24 h, the media were analyzed using LC–MS/MS.

Fig. 5
figure 5

CYP709C56 detoxifies MM through O-demethylation in vitro. A Immunoblotting of microsome fractions extracted from yeast expressing CYP709C56. B Proposed change in the structure of MM following O-demethylation catalyzed by CYP709C56. LC–MS/MS analyses of MM residue (C) and its O-demethylated metabolite (D) in yeast harboring the empty vector or expressing CYP709C56. Standard: analytical grade mesosulfuron-methyl and its O-demethylated metabolite. Vector control: mixture of yeast-expressed BSA protein and mesosulfuron-methyl. CYP709C56: mixture of yeast-expressed CYP709C56 enzyme and mesosulfuron-methyl. IA represents integral area

Full size image

Because the O-demethylation often occurred to sulfonylurea herbicides [26], we presumed that CYP709C56 could detoxify the herbicide MM via O-demethylation (Fig. 5B). Analysis of the reaction system showed that the content of MM detected in the media of yeast expressing CYP709C56 was much less than in the empty-vector control (Fig. 5C). In addition to MM, another peak with a shorter retention time, corresponding to that of the standard of O-demethylated MM was detected only in the presence of the CYP709C56 enzyme (Fig. 5D). Clearly, the CYP709C56 could metabolize MM through O-demethylation in vitro.

Docking analysis reveals structural interactions of CYP709C56 and MM

As indicated above, CYP709C56 catalyzes a reaction involving the O-demethylation of the methoxy of the pyrimidine ring from MM molecule (Fig. 5B). To gain further insights into the potential reaction mechanisms involved, a molecular docking approach was adopted. For this purpose, homology modeling was established for CYP709C56 protein based on the X-ray structure of human CYP4B1 (PDB ID: 6C93) (Fig. S3A). Similar to human P450 4B1 for which structure has been reported, CYP709C56 has multiple main alpha helix regions and two beta strands regions. The Ramachandran plot for CYP709C56 shows that 99% residues are in allowed regions (Fig. S3B), indicating that the three-dimensional structure of the model is reasonable. Structural analysis of the P450 modeling shows the CYP709C56 structure is basically consistent with the template structure (Fig. S3C). The average root-mean-square deviation value of the three-dimensional structure overlap is 1.935 Å, and both have the same alpha helix and beta strands regions.

Based on this model, substrate binding and specificity may involve the amino acid residues Thr-328, Thr-500, Asn-129, Gln-392, Phe-238, and Phe-242 (Fig. 6, Fig. S3D). For the MM molecule specifically, its three oxygen atoms, regarded as hydrogen bond acceptors, form three hydrogen bonds with the backbone nitrogen atom of Phe-242 and the side-chain nitrogen atoms of Asn-129 and Gln-392, respectively. The benzene ring of MM forms pi–pi stacking interactions with the benzene rings of Phe-238 and Phe-242, respectively (Fig. 6B). Mesosulfuron-methyl-CYP709C56 interaction energy in this case stabilizes at a level of − 4.76 kcal mol−1. In the current docking model, Thr-328 and Thr-500 are positioned close to the pyrimidine ring and the heme group, which may directly interact with the substrate methoxy for achieving its O-demethylation.

Fig. 6
figure 6

Molecular docking model for A. aequalis CYP709C56 with the molecule of MM. A The binding model of MM on the molecular surface of CYP709C56. MM is colored in cyan, heme is colored in purple, and the molecular surface of CYP709C56 is colored in pale green. B Close-up view of the substrate binding pocket derived from the CYP709C56 model. MM is colored in cyan, heme is colored in purple, the surrounding residues in the binding pockets are colored in yellow, and the backbone of the receptor is depicted as white cartoons with transparency

Full size image

Association between metabolic herbicide resistance and basal expression levels of CYP709C56

Mesosulfuron-methyl susceptibility in 48 populations of A. aequalis was compared in the absence and presence of malathion, a P450 inhibitor, to check whether P450s are involved in the MM tolerance. Seedling growth of seven populations (designated SD-02, SD-06, AH-15, AH-25, AH-29, AH-36, and JS-03) treated with MM in combination with 1 × malathion was significantly suppressed compared with growth without malathion (Fig. 7). On the contrary, seedling growth of the other 41 populations treated with MM was similar both with and without malathion pre-treatment (Fig. S4), indicating P450s are involved in MM tolerance in the above seven populations.

Fig. 7
figure 7

Effects of the P450 inhibitor malathion on the susceptibility of different populations to MM, along with the relative expressions of CYP709C56 in different populations. The expression of CYP709C56 in a specific population relative to that in the SD-01 population is shown by brown diamond. Bars indicate SE (n = 3), and *represents significant difference at P < 0.05

Full size image

MM metabolism was examined by LC–MS/MS in different populations of A. aequalis at 24 h after treatment. Compared with the S line SD-01, significantly lower levels of MM and correspondingly higher levels of O-demethylated metabolite were observed in plants of the above seven populations (Table 1). Therefore, P450s-involved enhanced rates of MM metabolism are present in these seven field-collected A. aequalis populations. Relative expression of CYP709C56 was then measured in the seven A. aequalis populations exhibiting metabolic herbicide resistance and compared with the S population SD-01. CYP709C56 significantly up-regulated (fold change ≥ 2, P < 0.05) in five of the seven populations (Fig. 7), exhibiting a high correlation with P450s-inbihitory phenotypes. However, in populations SD-02 and AH-36, constitutive expression of CYP709C56 was not significantly different with that in SD-01 (Fig. 7), suggesting it may not play a role in their metabolic resistance to MM.

Table 1 Mesosulfuron-methyl metabolism and target-site mutations in A. aequalis plants of field populations with obvious P450s-inhibitory phenotypes in comparison with a known herbicide-susceptible population (SD-01)
Full size table

Discussion

Since the late 1990s, P450s have been at the center of herbicide metabolism research because of their ability to endow selectivity in crops and resistance in weeds. However, the identification of genes encoding herbicide-metabolizing P450s is an arduous task in plants, as plants possess hundreds of P450s with varying substrate specificities. Despite the fact that the enhanced herbicide detoxification in weedy species has drawn great interest in recent years [35, 37], the underlying biochemical mechanisms, enzymes, and specific genes are still not well characterized and remain markedly under-explored. Potential candidates are intensively identified based on large-scale omics [38,39,40,41,42,43], whereas seldom of them have been demonstrated to be associated with specific herbicide metabolism [18, 44, 45]. In the weedy species A. aequalis, while metabolic resistance has been identified at the whole plant level for several years [46], little progress has been made in resistance gene discovery due to great genetic heterogeneity and lack of genome sequences and genetic linkage maps. Our previous studies have shown that the P450s are involved in the metabolic herbicide resistance of A. aequalis, and the up-regulated expressions of four P450 genes may be related to the resistance [31]. In this study, we characterize the functions of the candidate P450s and provide clear evidence that a novel P450 gene CYP709C56 metabolizes MM via O-demethylation and thus endows A. aequalis with herbicide resistance.

In some crops and grass weeds including Oryza sativa, E. phyllopogon, and Lolium rigidum, CYP81As are demonstrated to be involved in the concomitant resistance to multiple herbicides with different MOAs [13, 14, 45, 47]. Han et al. [45] hence proposed a theory of convergence evolution in P450-mediated NTSR. However, CYP81As seem not to be involved in the resistance of A. aequalis, at least in its resistance to MM. In comparison, CYP709C56 was suggested to play a key role in the resistance/tolerance of A. aequalis to specific ALS inhibitors. CYP709C56 is a member of the CYP72 clan, a quite diverse P450 family widely distributed in plant lineages and has undergone rapid functional divergence [48, 49]. Irmler et al. [50] demonstrated activity of the first member of the family, CYP72A1, as secologanin synthase in indole alkaloid biosynthesis. More recently, gibberellin-metabolizing activity was found in some members of CYP72As in the Brassicaceae lineage [51]. Notably, Imaishi and Matumoto [52] found that the rice CYP72A18 can metabolize a nine-carbon fatty acid structure herbicide pelargonic acid by (ω − 1)-hydroxylation, and Saika et al. [15] demonstrated the rice CYP72A31 can confer multiple resistance to the ALS-inhibiting herbicides bispyribac sodium and bensulfuron-methyl. For the CYP709 family specially, expression of several CYP709C genes is reported to be induced by treatment with different herbicides or safeners [53,54,55]. CYP709C1 expressed in yeast exhibits the potential for hydroxylating different fatty acids with varying chain lengths (C12–C18) and unsaturation [53], suggesting its possible detoxifying function. These previous reports together with our results suggest that CYP72 genes are widely involved in xenobiotic responses.

In this current study, CYP709C56 conferred tolerance to both the sulfonylurea herbicide MM and the triazolo-pyrimidine herbicide PX (Fig. 3A), which may be caused by the substrate specificity and amount of enzyme. It has been suggested that it would be difficult to predict substrate preferences from primary sequences in CYP72A proteins [48]. Here we built the homology modeling for CYP709C56 and characterized its structural interaction with the molecule of MM. Interestingly, the best fitting model structure presented by NCBI BLAST is human CYP4B1, which is a P450 gene functions in both endobiotic and xenobiotic metabolism [56]. P450s are heme-thiolate proteins able to catalyze the activation of molecular oxygen using the electrons from NADPH [24], and the reactions for P450-mediated herbicide metabolism primarily involve alkyl-hydroxylation, N-demethylation, O-demethylation, and aryl-hydroxylation including NIH-shift [26]. With the aid of critical cofactors of P450 reductase including flavin adenine dinucleotide and flavin mononucleotide, CYP709C56 most likely to utilize the two electrons (one hydride anion, H−) donated by NADPH to realize the O-demethylation of MM.

Transgenic Arabidopsis overexpressing A. aequalis CYP709C56 was resistant to MM and PX but was susceptible to FS. This resistance profile generally but does not completely in accordance with what is evident in the R A. aequalis, since the latter also confers moderate resistance to FS [46]. Besides, the resistance indexes determined in transgenic Arabidopsis for MM were much lower than that determined in R A. aequalis [31]. The distinct herbicide responses between the transgenic Arabidopsis and the field evolved A. aequalis support the hypothesis that the herbicide resistance in R A. aequalis population is caused by different mechanisms/metabolic genes. Previous RNA sequencing revealed that several other genes including GSTs, GTs, and ABC transporters also exhibited higher basal and/or herbicide-induced expressions in the R population [31]. Of them, GST genes GSTF1 and GSTU1 have been shown to be involved in the resistance of A. myosuroides and T. aestivum to specific herbicides, respectively [16, 57]. This suggests that the enhanced herbicide metabolism in R A. aequalis plants most likely to be regulated by multiple genes, at least including both P450s and potential GSTs, which is in accordance with well-studied cases of non-target-site-resistant weeds including L. rigidum and A. myosuroides [11, 12, 58,59,60]. The great potential of candidate GSTs in endowing A. aequalis plants with multiple herbicide resistance deserves greater research attention.

In this study, we clearly showed the role of CYP709C56 in metabolic herbicide resistance of the R A. aequalis population (AH-18). However, this gene is not consistently manifested among other resistant populations tested (Fig. 7). This is expected especially in genetically diverse, partly cross-pollinated weedy species A. aequalis [61], because the multiple metabolic genes/mechanisms will be selected within individuals and among populations [45], sometimes also diverse with the abiotic stresses, such as herbicide application history. This also corroborates the P450 and other genes can flexibly regulate the adaptive evolution of metabolic resistance in A. aequalis [45].

Identification of key genes conferring metabolic resistance to specific herbicides makes it possible to develop new biotechnology applications toward improving resistant-weed management. On one hand, basal expression levels of CYP709C56 may be used as molecular markers for screening putative resistant A. aequalis populations to confirm metabolic resistance to MM and/or PX. A similar approach has been utilized in the mosquito Anopheles funestus in which the first DNA-based metabolic resistance marker GSTe2 in mosquitoes was detected and developed as an essential tool to track the evolution of resistance [62]. In addition to improved resistance screening tools, the coding sequence of CYP709C56 could be utilized to synthesize new chemical inhibitors of herbicide-detoxifying P450s [37]. For instance, silencing of P450 genes in insects by RNA interference has been shown to increase insect susceptibility to insecticides [63, 64]. Besides, the coding sequence of CYP709C56 could also be used to engineer targeted gene knockout strategies to ultimately overcome the metabolic herbicide resistance. A pioneering work reveals that polynucleotide-based gene knockdown systems are being generated [65], which is especially suitable for overcoming metabolic herbicide resistance in weeds, as this is often under polygenic control.

Although we explicitly identify the CYP709C56 of considerable effect endowing metabolic herbicide resistance in A. aequalis, further study will illustrate how this gene is regulated. In this study, no difference was revealed in comparing sequences of CYP709C56 between the S and R plants (Fig. S1). Therefore, higher basal expression of CYP709C56 is not due to the difference in sequences but may be caused by mutations or varying degrees of methylation in the untranslated regions, gene promoters, or transcription factors [26, 66, 67]. Cloning of the promoter sequence and analysis of its methylation status, as well as identification of related transcription factors will likely reveal genetic regulation of the P450 gene [68,69,70]. While this study reveals the serendipitous role of CYP709C56 in herbicide detoxification, employing metabolomics will aid in understanding the endogenous function of CYP709C56 in specialized metabolism in plants [18]. To reveal genetic architecture of P450-driven evolution of metabolic resistance in A. aequalis, genome-based work is needed to conduct.

Experimental procedures

Plant materials and growth conditions

The collection, storage, and planting of seeds from the S (SD-01) and R (AH-18) A. aequalis populations were the same as described elsewhere [31]. Fresh leaves were collected from A. aequalis seedlings at the 3–4-leaf stage, and total RNA was extracted using the EasyPure® Plant RNA Kit (Trans, Catalog No. ER301). Reverse transcription (RT) was conducted with 5 μg of total RNA as template using the TransScript® II One-Step gDNA Removal and cDNA Synthesis SuperMix (Trans, Catalog No. AH311).

Arabidopsis (Arabidopsis thaliana, ecotype Columbia-0) was used as a model species in this study. All Arabidopsis plants were grown in a controlled-environment growth chambers setting at temperatures of 25/20 °C with a long-day condition (16-h light/8-h dark) for bolting or with a short-day condition (8-h light/16-h dark) for vegetative growth, unless specified otherwise.

Vector construction and plant transformation

The full-length coding region of CYP709C56 was amplified by RT-PCR with the primer pairs listed in Table S2 using cDNA prepared from fresh shoots of A. aequalis at the 3–4-leaf stage. Sequence alignment was performed by Clustal W [71] and ESPript v.3.0 [72]. Phylogenetic analysis was conducted using MEGA-X [73]. Physico-chemical parameters were computed by the Expasy’s ProtParam prediction server (http://us.expasy.org/tools/protparam.html).

The amplicons were subcloned into the pEASY®-T1 cloning vector (Trans, Catalog No. CT111) to yield entry vectors. The plasmids were jointly digested by QuickCut™ KpnI (TaKaRa, Catalog No. 1618) and QuickCut™ BamHI (TaKaRa, Catalog No. 1605), and then ligated to the binary vector pPZP211 [74] which carries the CaMV 35S promoter. The binary vector was transformed into chemically competent Agrobacterium tumefaciens strain GV3101 through freeze–thaw method [75]. The transformed Agrobacterium strains were used to transform the Arabidopsis with the floral dip method [76]. The transformants were selected by kanamycin treatment at 50 mg L−1. Same procedures were conducted to obtain the homozygous Arabidopsis transformants overexpressing CYP71R18, CYP94C117, and CYP94E14, respectively.

Herbicide susceptibility of transgenic Arabidopsis

Herbicide susceptibility of the transgenic lines was evaluated by growth on MS media containing different concentrations of each herbicide. Three classes of ALS inhibitors including the sulfonylurea herbicide MM, the triazolo-pyrimidine herbicide PX, and the sulfonyl-aminocarbonyl-triazolinone herbicide FS were used for the testing. All assays were conducted at 25/20 °C with short-day conditions and a light intensity of 200 µM m−2 s−1. After sterilization and washing, the seeds of transgenic Arabidopsis were planted on MS media containing different concentrations of each herbicide in petri dishes and allowed to grow for 14 days [16]. The dishes in the growth chamber were re-arranged every other day. At 14 days after treatment, susceptibility of transgenic Arabidopsis to different herbicides was visually assessed.

The MM susceptibility of independent transgenic lines was determined by spraying whole plants [16]. At 9 days after treatment, the Arabidopsis seedlings were cut at ground level, and their aboveground weights were determined and expressed here as a percent of untreated controls. Each harvest contained three replications, and the GR50 was calculated by fitting the data to a four-parameter log-logistic curve using SigmaPlot v.14.0 (Systat Software, San Jose, CA, USA) [77].

RT-qPCR assay

Individual 21-day-old seedlings of T3 homozygous lines were harvested for RNA extraction. RT-qPCR was carried out using the TransStart® Tip Green qPCR SuperMix (Trans, Catalog No. AQ141) as described recently [78], and the primers used are listed in Table S2. mRNA quantitation was performed by the 2−∆Ct method [79]. The expression of target genes was normalized to that of the Arabidopsis housekeeping gene AtF-box protein (Accession no. At5g15710) [80].

LC–MS/MS analysis of MM metabolism in A. aequalis and transgenic Arabidopsis

Seedlings of A. aequalis and Arabidopsis at the suitable leaf stages were, respectively, used for MM metabolism analysis following foliar-spray with a commercial formulation of MM at 9 g ai ha−1 (30 g L−1 oil-miscible flowable, Sigma). The aboveground shoots of each line were harvested at different times after herbicide treatment. Each harvest had three replicate samples of at least five plants and these were snap-frozen in liquid N2 and stored at − 80 °C. The details of MM extraction and LC–MS/MS analysis by Xevo TQ-S (Waters) were the same as described previously [33]. MM and its O-demethylated metabolite were detected in negative-ion mode with m/z 502.10 and m/z 488.10, respectively. Statistical analysis of the data was conducted using the Student’s t test (P < 0.05) procedure in SPSS software (IBM, Armonk, NY, USA).

Yeast transformation and CYP709C56 expression

For the expression of CYP709C56, the WAT11 yeast strain and the pYeDp60 vector system were used [36]. The full-length coding sequence of CYP709C56 was amplified and integrated into the KpnI site of the pYeDp60 vector using the pEASY®-Uni Seamless Cloning and Assembly Kit (Trans, Catalog No. CU101). A triple-Flag epitope sequence was fused just downstream of the ATG through gene synthesis by Sangon Biotech. The recombinant plasmid was transformed into WAT11 using a LiAC method according to the manufacturer’s instructions (Takara, Catalog No. 630439). Yeast cells were induced according to the modified two-stage cultivation method using modified SLI medium [81]. Microsomal fractions were prepared according to Renault et al. [82], after which they were electrophoresed on 12% (w/v) sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Immunoblotting analysis was performed according to Iwakami et al. [13] using the Anti-Flag-Tag mouse monoclonal antibody (Sangon Biotech, Catalog No. D190828) as primary antibody.

Yeast expression of CYP709C56 was performed according to Iwakami et al. [14]. At 14 h after the induction of P450 by galactose, MM was added to the culture media at a final concentration of 60, 100, or 200 µM. After shaking for 24 h, the culture solution was centrifuged at 14,000g for 5 min. The supernatant was filtered through a 0.22-µm membrane and analyzed by LC–MS/MS (Waters) as described recently [33].

Resistance identification and CYP709C56 expression in A. aequalis

In May 2015 to May 2018, mature seeds of A. aequalis were collected from independent wheat fields across the Provinces of Jiangsu (JS), Anhui (AH), Henan (HN), and Shandong (SD), China. A total of 48 populations of A. aequalis were selected and used to test the susceptibility to MM with or without P450 inhibition (Figs. 7, S4). Weed seedlings at the 3–4-leaf stage were treated with malathion, MM, or malathion plus MM [30]. Malathion at 1000 g ai ha−1 (1×) and MM at 9 g ai ha−1 (1×) were used for the testing. At 21 days after treatment, the aboveground dry weight was determined and expressed as a percentage of untreated control. Plant susceptibility to herbicide was defined according to the following criterion: “R” plants displayed strong regrowth as the untreated control, “r” plants showed visible regrowth but much weaker, and “S” plants showed leaf chlorosis and desiccation. For the populations exhibiting significant P450s-inhibitory phenotypes, herbicide metabolism and relative expression of CYP709C56 were measured according to the methods described above and compared with the S population (SD-01).